More consistent way and node traversal

[ansoncfit]

Follow-up to #841 -- we may want to check the isImpassable method there against TraversalPermissionLabeler.isNoThruTraffic.

Example to test; the OSM ways for cars along Hollywood Bowl Rd (e.g. https://www.openstreetmap.org/way/158587690) have tag access=customers. isNoThruTraffic does not allow through-traffic on these ways, but origins can be linked to them? In this case, one solution may be tagging the entrance and exit gates with access=customers in OSM, and add this tag to the isImpassable method. Or along with #806, we can also look into more refined island pruning to handle cases like this.

[abyrd]

Discussed in today's meeting: our modeling of impassable barriers such as emergency exits is mainly to improve handling of bad OSM input data through island pruning. In the case of no-through-traffic areas we don't want to block them off completely because they are legitimate to traverse if the origin or destination point is closer to them than to any other road. So handling no-through-traffic areas is probably a separate routing issue. It requires blocking traversal based on accumulated state (once you've entered a thru traffic area you can't exit it, but you can exit it if you started in one) so this may violate some routing assumptions. There should be examples of how to handle this in OpenTripPlanner.

[ansoncfit]

Possibly related: access restrictions in the vicinity of https://www.openstreetmap.org/way/502174807